Skip to main content Skip to footer

Going Beyond the “Green Fuzzy”: Measuring Runoff Reduction at Modular Vegetative Roofs

May 15, 2011

S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1 GR A N T • 7 3
400F Cowgill Hall (0205), Blacksburg, VA 24061
Phone: 540-231-0609 • Fax: 540-231-9938 • E-mail:
7 4 • GR A N T S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1
Vegetative roofs’ capacity to decrease the rate and volume of roof runoff is dependent on
a range of factors related to climate and roof system characteristics. An experimental setup
constructed by the author at Virginia Tech is testing the effects of system depth and presence
of vegetation on runoff reduction. Data collection began in April 2011, and results from
selected storms will be presented. The expected outcome of the study is a model that will
enable vegetative roof system designers, building owners, and policymakers to more confidently
anticipate the benefits of installing modular vegetative roof systems for stormwater
ELIZABETH GRANT is an assistant professor at the School of Architecture + Design at
Virginia Tech. She holds a BA (1995), an MS in architecture (2003), and a PhD in architecture
and design research (2007)–all from Virginia Tech. She is a registered architect, a member
of RCI, Inc., and the 2004 recipient of the National Roofing Foundation’s William C.
Cullen Research Fellowship Award. She has published in Interface, the Journal of
Architectural Engineering, and the Journal of Green Building. Her interests include environmentally
sensitive design, the building envelope, and building systems integration.
The capacity of vegetative roofs to
decrease the rate and volume of roof runoff
is dependent on a range of factors related to
climate and roof system characteristics. An
experimental setup constructed by the
author at the Research and Demonstration
Facility at Virginia Tech’s College of Architecture
and Urban Studies is testing the
effects of system depth and presence of vegetation
on runoff reduction. Five 8-ft x 8-ft
(2.4-m x 2.4-m) test plots have been built to
simulate low-slope roofs. Three plots are
filled with prevegetated modular roof systems
of differing depths. A fourth plot contains
only growing medium, and a fifth is a
control. Rain gauges connected to downspouts
from each test plot are being used to
measure runoff, while a sixth rain gauge
measures rainfall. A weather station simultaneously
measures ambient air temperature,
relative humidity, solar radiation, and
wind speed and direction. Data collection
began in April 2011, and results from
selected storms are presented. The expected
long-term outcome of the study is a model
predicting runoff reduction for modular vegetative
roof systems. This model will enable
vegetative roof system designers, building
owners, and policymakers to more confidently
anticipate the benefits of installing
modular vegetative roof systems for
stormwater mitigation.
In the roofing trade, claims about the
benefits of new technologies tend to precede
their thorough vetting in the field. Failures
of first-generation green roofs in North
America are a testament to this phenomenon
(Osmundson, 1999). One of the solutions
to this problem is to import technology
that has been proven elsewhere and
implement it in a new context. Vegetative
roofs have largely followed this trajectory,
with many North American companies partnering
with German organizations with
years of experience. While this tactic may
have taken some of the risk out of green
roofing, the performance benefits that have
been proven overseas remain to be tested
locally. Several institutions, both commercial
and academic, have taken on this burden
of proof through research projects that
attempt to give an unbiased comparison of
green roof systems to one another and to
more typical roof configurations. The widely
touted benefits of green roofs include runoff
retention, pollutant mitigation, reduction of
heat flux through the roof assembly,
acoustic protection, and air quality
improvement. This paper will address perhaps
the most compelling and defensible
benefit of green roofs: their ability to reduce
roof runoff as compared to traditional roofing
There have been multiple studies conducted
in North America documenting the
ability of green roofs to reduce roof runoff.
Results of these studies vary greatly with
local weather, the time period studied, and
system characteristics; but the general consensus
has been that green roofs retain
somewhere between half and all of the precipitation
incident upon them. Because this
range is so wide, further research is needed
to more closely predict results. Another critical
factor related to runoff mitigation is
peak flow reduction, which helps to reduce
erosion and the overtaxing of storm sewer
systems that can result in urban flooding.
Many of the studies conducted in North
America have aimed to identify variables,
including the time of year, the interval
between storms, and the intensity and
duration of storms, to determine their
effects on the capacity of green roofs to
retain and delay stormwater. Architectural
factors such as roof slope, distance between
drains, the depth and makeup of green roof
medium, and the variety of vegetation
selected also have an influence on the success
of green roofs as stormwater management
devices. Efforts are ongoing to establish
the influence of each of these parameters
and provide a reliable, useful measure
of performance that may be used to quantify
the impact of green roofs in stormwater
The research installation described in
this paper is located at the College of
Architecture and Urban Studies’ Research
and Demonstration Facility on the campus
of Virginia Tech in southwest Virginia. A
modular green roof system was chosen for
study because of its local availability and
recent implementation as the first residential
green roof project in Blacksburg. The
system was easily installed by a team of
students without special equipment or
expertise. The system used is composed of
1-ft x 2-ft (0.3-m x 0.6-m) interlocking black
plastic trays of varying depths: a “Deep”
system with 6 in (152 mm) of medium, a
“Standard” system with 4.25 in (108 mm) of
medium, and a “Lite” system with 2.5 in (64
mm) of medium. The medium is a proprietary
blend of inorganic and organic components
sourced locally; in this case, the inorganic
component is rotary-kiln-expanded
slate lightweight aggregate. The blend contains
approximately 94% inorganic material
by dry weight and conforms to German FLL
granulometric standards. A fully saturated
Standard module weighs between 27 and
29 pounds per sq ft (130 and 140 kg per sq
m) when fully vegetated.
It was hypothesized that the medium
alone would have some runoff retention
properties; to test this hypothesis, a set of
Standard modules was filled with medium
and not planted. The modules with plants
Sedum kam ellacombium
Sedum album murale
Sedum stefco
Sedum spurium ‘Fuldaglut’
Sedum spurium ‘John Creech’
Sedum rupestre ‘Angelina’
Sedum sexangulare ‘Utah’
Table 1 – Plant list for vegetated
S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1 GR A N T • 7 5
were fully vegetated when delivered to the
research site, and all were planted with the
same mix of species listed in Table 1. The
bottoms of all of the modules are articulated
and slotted for drainage, and no filter
fabric was included.
Five platforms were constructed on the
roof of the one-story Test Cell Building,
immediately to the south of the main
Research and Demonstration Facility. The
Test Cell Building was
purpose-built for student
and faculty experiments.
The existing roof
is covered with concrete
pavers. This location
allowed for easy transfer
of the modules from
grade to roof.
Platforms were constructed
of dimensional
lumber and plywood in
lieu of placing the modules
on the existing roof
surface. This was done
for several practical reasons.
First, it was necessary
to provide enough
vertical clearance to
attach a standard gutter
to a downspout running
to a tipping bucket to
catch and record the
amount of runoff.
Attaching gutters to
each platform avoided
the difficulty of modifying
the building’s existing
concealed gutter to
isolate drainage for each treatment condition.
Second, elevating the experimental
platforms avoided potential damage to the
existing roof system. Third, the platforms
permit relatively simple removal or relocation
of the project at the end of the study
period, if necessary. Fourth, the area under
the platforms allows access to the underside
of the roof assembly, which may prove
useful in future thermal evaluations of the
vegetative roof systems. Finally, other
design projects coexist with this experiment,
and the platforms’ rooftop location
puts them at lower risk of accidental or
deliberate disruption.
The 8-ft x 8-ft (2.4-m x 2.4-m) platforms
were spaced 8 ft (2.4 m) apart on the roof to
ensure that their supports rest on the bearing
walls of the building below and to allow
for access on three sides for installation,
observation, and irrigation of the plants.
The setup is shown in Figure 1.
The platforms were built and protected
with tarps on July 6, 2010, and covered
Figure 1 – Research setup (base plan courtesy of Bill Galloway).
Figure 2 – Waterproofed platforms awaiting vegetated module installation.
7 6 • GR A N T S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1
with 45-mil TPO membrane
on August 11, 2010.
In a typical vegetative roof
installation, this membrane
would have been
covered with a slip sheet,
but here this was omitted
due to the relatively short
duration of the project and
the absence of foot traffic
on the platforms.
Membrane flashing was
adhered to solid aluminum
angles on the three
nondraining sides of the
platforms to form watertight
curbs, as shown in
Figure 2. The platforms
were leveled to slope
toward the gutters at ¼ in
per ft (1:48).
On October 12, 2010,
the vegetated modules
were brought to the site on
a nursery truck, driven,
and lifted to the rooftop via
forklift as shown in Figure 3 and installed
by student volunteers as shown in Figure 4.
The draining sides were fitted with slotted
angles to help hold the modules in place
while allowing runoff to reach the gutters.
A Campbell Scientific weather station
was erected on the west side of the roof,
comprising the following equipment: a
Vaisala HMP50 temperature and relative
humidity probe, a LI-COR LI90SB photosynthetically
active radiation sensor, a RM
Young 03002 wind sentry set measuring
wind speed and direction, and a Hydrological
Services TB6 tipping bucket rain gauge. A
Campbell Scientific CR1000 data logger in a
weatherproof enclosure was attached to the
weather station’s tripod. The data logger
Figure 3 – Modules
brought to platforms
via forklift.
S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1 GR A N T • 7 7
Figure 4 – Student
volunteers installing
records signals from the above-mentioned
sensors along with tipping buckets at each of
the five platforms and CS616 soil moisture
sensors at the three vegetated platforms.
Data are downloaded to the researcher’s laptop
every week during site inspections.
Hydrological Services TB1L tipping
buckets with 0.5-L tipping volumes were
hooked to the end of short lengths of downspout
at each of the five platforms. These
large tipping buckets were chosen because
they can handle the considerable amount of
runoff from a 64-sq-ft (5.9-sq-m) area during
a heavy storm while also being three
times as sensitive, on a rainfall-perarea
basis, as the TB6 rain gauge
used to record rainfall. Because of
the relatively large tipping volume of
the TB1L, residual water sitting in
the bucket between storms may
slightly skew the data for the completed
storm, or for the next storm
in the data set. Because it was unrealistic
to travel to the research site
to empty and reset the tipping buckets
prior to every storm, this is a
potential source of error in the
reported data.
Like any roof, this project cannot
be left unattended. Weekly visits
are necessary to examine the condition
of the tipping buckets, check on
the health of the plants, and download
the data. While green roofs are
designed to be largely self-sustaining,
watering is required during
warm periods of drought exceeding
two weeks and, in the height of
summer, one week. Knowing when
plants must be watered is, at present, more
of an art than a science, and an interesting
future research direction would be to develop
an algorithm to direct manual or automatic
irrigation schedules for green roofs.
For this project, the watering regimen has
been conducted according to the advice of
Date rainfall Total Rainfall
recorded rainfall (mm) equivalent (L) Runoff per platform (L)
(at 0:00)
#1 #2 #3 #4 Std. #5
Deep Standard Lite Med. Control
6/6/2011 5.3 31.8 0.5 1.0 0.5 1.0 27.5
6/8/2011 0.3 1.5 0.0 0.0 0.0 0.0 3.0a
6/11/2011 25.7 152.9 56.5 19.0 88.0 55.0 137.0
6/13/2011 7.6 45.4 1.0 1.0 3.0 1.5 40.0
6/14/2011 0.3 1.5 0.0 0.0 0.0 0.0 0.5
6/19/2011 14.0 83.3 5.0 4.0 16.0 15.5 69.5
6/21/2011 1.3 7.6 0.0 0.0 0.5 0.0 7.0
Total for period 54.4 324.0 63.0 24.5 108.0 73.0 284.5
Total % of rainfall retained 81 92 67 77 12
Table 2 – Runoff and rainfall data for seven storms from June 1 to June 21, 2011
7 8 • GR A N T S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1
Figure 5 – Vegetated platforms after pruning on June 13, 2011. The Deep system is in
the foreground.
aThe runoff from platform #5 is greater than the rainfall due to the greater sensitivity of the tipping bucket at the platform versus the rain gauge, and also likely due to
residual water left in the tipping bucket from the previous storm.
the landscaping firm supplying the vegetated
modules. Plants were hand-watered with
a garden sprayer at a rate of 1.25 gallons
(4.73 liters) per module upon installation,
with additional irrigation on November 9,
2010, and June 2, 2011.
The spring of 2011 was rainy, which
meant little irrigation was needed, and the
plants thrived. Because of this, the landscape
supplier recommended that the vegetation
be cut back to avoid a humid environment
with low light that can harbor fungal
infections such as botrytis. All three vegetated
platforms were lightly pruned on
Monday, June 13, 2011. Cuttings were left
on platforms #2 and #3 to help fill in small
bare spots but removed from the fully covered
platform #1, again at the advice of the
landscape supplier. Though no formal evaluations
were made, a positive correlation
was observed between the depth of the
medium and the vigor, size, coverage, and
height of the plants. Figure 5 shows the
platforms after the modules were pruned,
with the most aggressive growth on the
Deep system in the foreground.
Data collection began in April 2011; and
after an initial period of troubleshooting,
flow data were recorded starting on June 1,
2011. Results from June 1 through June 21
are presented in this paper. Seven total rain
events occurred during this period. Rainfall
events were combined when rain from a
second storm occurred within six hours of
rain or runoff from the first storm.
Tabulated results for the study period are
presented in Table 2. “Rainfall equivalents,”
or the amount of rainfall incident on each of
the platforms in liters, was estimated by
taking measurements from the TB6 rain
gauge and extrapolating these to a 64-sq-ft
(5.9-sq-m) platform. The final row of Table 2
gives the percentage of rainfall retained by
each platform for the study period, calculated
as follows:
(1 – Total runoff per platform for period
/Total rainfall equivalent for period)
x 100
Figure 6 shows the amount of roof
runoff retained on each platform as a percentage
of total incident rainfall on that
platform during each storm event, to illustrate
the difference in performance on a
per-storm basis. The two storms in the
reported period with less than 0.05 in (1.3
mm) of rainfall are not graphed in Figure 6,
since the four treatment platforms showed
no runoff during these two events.
The small sample size makes sweeping
conclusions premature, but several trends
have emerged. First, it was clear that all of
the vegetated platforms and the mediumonly
platform retained considerably more
runoff than the control platform. Second, it
was apparent that the greater the amount
of precipitation, the greater the difference
between the treatment platforms and the
control platform, as seen in Figure 6. In
lighter storms, the amount of runoff from
the treatment platforms became almost
negligible. Third, the differences among the
vegetated treatments–Deep, Standard, and
Lite–were somewhat inconclusive. The Lite
system underperformed the Deep and
Standard systems in four of the five storms
with rainfall totals of 0.05 in (1.3 mm) or
more, but the Deep system did not consistently
outperform the Standard system, as
might have been expected. In fact, the total
runoff retained by the Standard system
(92%) exceeded that retained by the Deep
system (81%) for the reported period.
Fourth, the Standard medium-only system
yielded somewhat surprising results. It
retained more runoff than the Lite system
in four out of the five storms with 0.05 inches
(1.3 mm) or more rainfall. Precedent for
this outcome was found by VanWoert et al.
(2005) who reported no significant difference
between vegetated and medium-only
treatments when storms were grouped by
intensity, leading them to suggest that the
characteristics of the growing medium and
(in their case) the retention fabric have the
greatest effect on runoff reduction. Fifth,
the control platform retained 12% of the
total rainfall for the studied period, which is
higher than expected for a membrane roof
and may be due to the differing sensitivities
of the tipping bucket at the control platform
versus the rain gauge. As data collection
progresses, further investigation with a
larger number of storms may clarify the
preliminary findings presented here and
allow for meaningful interpretations.
S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1 GR A N T • 7 9
Figure 6 – Roof runoff retained as a percentage of total incident rainfall for five storms from June 1 to June 21, 2011,
where #1 is Deep, #2 is Standard, #3 is Lite, #4 is Standard medium only, and #5 is Control.
To develop a fuller picture of modular
vegetative roofs’ contribution to stormwater
mitigation, reduction of peak flows will be
calculated when a significant number of
heavy storms have been recorded. The
interrelationships among temperature,
humidity, solar radiation, wind speed, soil
moisture, storm intensity, time between
storms, and green roof runoff reduction will
also be explored as this project progresses.
The study will be continued for a full year,
after which analysis of the data will support
construction of an algorithm useful for predicting
future performance of modular vegetated
roofing systems. Upon completion of
this phase of the study, the platforms may
be modified to measure the influence of
other variables, such as roof slope, on green
roof performance. Further, pollution mitigation
may be investigated, along with thermal
impacts of vegetative roofs, as the
research agenda expands.
The author would like to acknowledge
the following partners of Virginia Tech’s
Center for High Performance Environments
for their support of this research: RCI, Inc.,
the National Roofing Contractors
Association, and Riverbend Nursery.
Thanks also go to GAF, Acrylife, and
Systems Construction LLC for their in-kind
donations of time and materials, and to Dr.
Brad Rowe of Michigan State University for
his helpful guidance.
T. Osmundson. Roof Gardens: History,
Design, and Construction (1st ed.).
New York: W.W. Norton, 1999.
N.D. VanWoert, D.B. Rowe, J.A.
Andresen, C.L. Rugh, R.T. Fernandez,
and L. Xiao. “Green Roof
Stormwater Retention: Effects of
Roof Surface, Slope, and Media
Depth.” Journal of Environmental
Quality, 34, 1036-1044, 2005.
8 0 • GR A N T S Y M P O S I U M O N B U I L D I N G E N V E L O P E T E C H N O L O G Y • OC T O B E R 2 0 1 1